AI Usage Policy for Small Business: Rules and Free Template
By Nemanja Vukmirović, Lizzard Solutions. Last updated 15 November 2026.
Someone in your sales team pastes a client's full contract into a free AI chatbot to get a quick summary. Someone in accounting uploads a spreadsheet with salaries to build a chart. Nobody did anything malicious. But nobody knows where that data went, either, because nobody ever wrote down the rules.
An AI usage policy for a small business fixes that. It does not need to be a 30-page legal document. One or two pages that say which tools are allowed, what data never goes into them and who checks the output are enough to remove most of the risk. This guide explains what to include and gives you a template you can adapt today.
TL;DR
- An AI usage policy is a short internal document that sets rules for how employees use tools like ChatGPT, Gemini, Copilot or Claude.
- The core of it: an approved tool list, a data traffic-light (what may and may not be pasted), human review of outputs, and one owner.
- Paid business plans and free personal accounts often handle your data differently, so the tool list matters.
- If you operate in the EU, the AI Act already expects staff using AI to understand its risks. A short policy plus a training record is a practical start.
What is an AI usage policy?
An AI usage policy is a written set of rules that tells employees which AI tools they may use for work, what information they may share with them, how they must check AI output and who to ask when unsure. It sits next to your IT security policy and your data protection rules, and it applies to everyone: employees, interns and contractors who work on your behalf.
Without one, every person makes their own rules. Some refuse to touch AI and lose time. Others use it for everything, including things that should never leave the company.
Why a small business needs one now
- Data leaks are the real risk. Client data, contracts, passwords and source code pasted into a personal AI account may be stored or processed under terms your company never agreed to.
- AI output can be confidently wrong. Invented facts, wrong legal references and broken code look just as polished as correct ones.
- Clients are starting to ask. Larger clients increasingly include questions about AI use in supplier questionnaires.
- Regulation is arriving. Under Article 4 of the EU AI Act, companies that use AI systems must take measures to support the AI literacy of their staff. The European Commission's AI literacy Q&A says this already applies, and gives a direct example: employees who use ChatGPT to write ads or translate text should be informed about risks such as hallucination.
The same Commission page notes that no certificate is required and that an internal record of trainings and guidance is acceptable. For a small company, that makes a short policy plus a signed acknowledgement a sensible, low-cost step. This is not legal advice: if you serve EU clients or have EU operations, check your obligations with a lawyer.
Key takeaway
You do not need to ban AI or buy an expensive governance platform. You need clear, written rules that people actually read.
The 7 parts of a practical AI usage policy
- Purpose and scope. One paragraph: why the policy exists and who it covers.
- Approved tools. A named list, with the account type (for example a company business plan, not personal free accounts). New tools need approval before use.
- Data rules. What may never be entered, what needs approval, and what is fine. The traffic light below makes this easy.
- Human review. AI drafts, people decide. Anything sent to a client, published or used in code is checked by a person who is responsible for it.
- Transparency. When AI use must be disclosed, for example to clients, in published content or in automated chats with customers.
- Security basics. Company accounts with multi-factor authentication, no shared logins, and no connecting AI tools to email or drives without approval.
- Owner, training and review. One named person answers questions, runs a short training, keeps the acknowledgement record and reviews the policy every six months.
The data traffic light
Most people will not memorise a policy. They will remember three colours. Put this table in the policy and on the wall.
| Colour | Examples | Rule |
|---|---|---|
| Red: never | Passwords, API keys, ID numbers, health or salary data, full client contracts, unreleased financials, source code under NDA | Never paste into any AI tool, even approved ones, unless a specific exception is approved in writing |
| Yellow: approved tools only | Internal documents, anonymised client info, company procedures, non-sensitive code | Only in approved business accounts; remove names and identifiers first |
| Green: free to use | Public information, marketing drafts, generic questions, your own writing to polish | Any approved tool; still review the output before using it |
Common mistake
Writing "do not share confidential data" and stopping there. People do not agree on what "confidential" means. Give concrete examples in each colour.
Personal data deserves extra care. If a prompt contains names, emails or phone numbers of clients, your data protection obligations still apply. Our guide to the Serbian data protection law covers the basics for Serbian companies.
AI usage policy template (copy and adapt)
Replace the parts in brackets. Keep it short. A policy that fits on two pages gets read.
Template
1. Purpose. [Company] supports the use of AI tools to save time and improve quality, as long as client and company data stay protected. 2. Scope. Applies to all employees, interns and contractors working for [Company]. 3. Approved tools. [Tool + business plan], [Tool + business plan]. Using any other AI tool for work requires approval from [owner]. 4. Data. Follow the traffic light: red data never goes into AI tools; yellow only into approved tools and anonymised; green is free to use. 5. Review. Every AI output used for clients, publication, decisions or code is checked by a responsible person. AI is never the final author. 6. Disclosure. We tell clients when AI is used in work delivered to them where [contract / client] requires it, and customers when they talk to an AI chatbot. 7. Accounts. Use company accounts with MFA. Do not connect AI tools to email, drives or the CRM without approval. 8. Incidents. If red data was entered by mistake, report it to [owner] the same day. 9. Owner and review. [Name] owns this policy, runs onboarding training and reviews it every six months. Signed by employee: ____ Date: ____
Add the policy to your employee onboarding IT checklist so every new hire reads and signs it on day one.
Example: a 12-person accounting office
An accounting office notices that staff use several free AI chat apps to draft client emails and explain tax rules. Some prompts contain client names and income figures. The owner does not ban AI, because it genuinely saves time on routine emails.
Instead, the office picks one tool on a business plan, writes a two-page policy with a traffic light (client financial data is red, anonymised questions are yellow, email tone and grammar are green), and adds one rule: AI answers about tax law are never sent to a client without checking the official source. A 30-minute training covers real examples, and everyone signs. The owner keeps the signed sheets as the training record.
The result is not zero risk. It is known, documented risk with a clear owner, which is what clients and auditors want to see.
Beyond chat: AI in your products and automations
If you go further and connect AI to your systems, such as a customer-facing AI chatbot or AI agents that act in your tools, the policy needs a technical layer too. The OWASP Top 10 for LLM applications lists risks like prompt injection and sensitive information disclosure, and the NIST AI Risk Management Framework offers a structured, voluntary approach to managing AI risk. A small business does not need to implement either in full, but both are useful checklists when you build.
Not sure where to start? Our AI readiness checklist helps you decide which processes are worth automating, and our overview of AI integration for small business explains how projects typically run.
FAQ
Should a small business ban ChatGPT?
Usually no. A ban tends to push people to personal accounts on their phones, where you have no control at all. Approving a tool and setting data rules gives you more protection.
Is a free AI account safe for company data?
It depends on the provider's terms, which can differ between free, paid and business plans and can change. Read the data use and retention terms for the exact plan, and treat free personal accounts as unsuitable for yellow or red data.
How long should an AI usage policy be?
One to two pages is enough for most small companies. Add longer technical rules only when you build AI into products or automations.
Do employees need AI training?
Yes, at least a short session with real examples of what can go wrong, such as invented facts and data leaks. In the EU, Article 4 of the AI Act expects staff to have adequate AI literacy, and an internal record of that training is acceptable according to the European Commission.
Want AI in your business, without the risk?
We help small businesses choose AI tools, set clear usage rules and build safe automations. Tell us how your team works and we will propose a practical setup.
Talk to us